Scott Feltman
Research Subjects
Scott Feltman has a total of 30 co-authors. Here are the top co-authors by number of shared publications:
- Roman Kotov
Department of Psychiatry, Stony Brook University , Stony Brook, NY 11794
7 shared publications Benjamin Luft
5 shared publications- Whitney R. Ringwald
Department of Psychology, University of Minnesota
5 shared publications - Adithya V. Ganesan
Department of Computer Science, Stony Brook University, Stony Brook, USA
4 shared publications - H. Andrew Schwartz
Department of Computer Science Stony Brook University Stony Brook New York USA
4 shared publications - Oscar Nils Erik Kjell
Department of Psychology, Lund University, Lund, Sweden
4 shared publications
Latest Publications
- Open Access
- 6 November 2026
- Published by Center for Open Science in PsyArXiv
Do thoughts actually drive symptoms? Testing cognitive theories of posttraumatic stress with natural language processing in daily diaries
Cognitive theories are among the leading conceptualizations of posttraumatic stress disorder (PTSD). The core perspective shared by these theories is that distorted thinking (e.g., “no one can be trusted”) is involved in maintaining symptoms. However, the proximal mechanisms linking cognition and PTSD symptoms remains unclear due to two major methodological limitations: 1) lack of research on cognition-symptom dynamics at the daily level, and 2) reliance on self-reported thought distortions, which is a method ill-suited for capturing biases that people often do not have insight into or are outside of their awareness. The present study helps address these limitations in a sample of trauma-exposed patients (N=510) who completed daily video diaries and symptom ratings for 90 days (n=30,958 days). We used two natural language processing techniques to assess cognitive distortions from the diary transcripts: a lexica-based and large language model (LLM) approach. We then established the day-to-day sequencing of cognitive distortion and symptom fluctuations using dynamic structural equation models. Across the lexical and LLM models, distorted thinking did not predict worsening symptoms (bs=-.02–.01). Instead, worsening symptoms predicted more distorted thinking (bs=.03–.07). Also, symptoms and cognitive distortions tended to fluctuate together (bs=.08–.41). The current findings suggest that cognitive distortions do not precede PTSD symptoms as many theories have assumed, but may instead be a consequence or correlate of them. If replicated, our results could help refine cognitive and related models of PTSD.
- 6 November 2026
- Published by Center for Open Science in PsyArXiv
Open AccessDo thoughts actually drive symptoms? Testing cognitive theories of posttraumatic stress with natural language processing in daily diaries
Cognitive theories are among the leading conceptualizations of posttraumatic stress disorder (PTSD). The core perspective shared by these theories is that distorted thinking (e.g., “no one can be trusted”) is involved in maintaining symptoms. However, the proximal mechanisms linking cognition and PTSD symptoms remains unclear due to two major methodological limitations: 1) lack of research on cognition-symptom dynamics at the daily level, and 2) reliance on self-reported thought distortions, which is a method ill-suited for capturing biases that people often do not have insight into or are outside of their awareness. The present study helps address these limitations in a sample of trauma-exposed patients (N=510) who completed daily video diaries and symptom ratings for 90 days (n=30,958 days). We used two natural language processing techniques to assess cognitive distortions from the diary transcripts: a lexica-based and large language model (LLM) approach. We then established the day-to-day sequencing of cognitive distortion and symptom fluctuations using dynamic structural equation models. Across the lexical and LLM models, distorted thinking did not predict worsening symptoms (bs=-.02–.01). Instead, worsening symptoms predicted more distorted thinking (bs=.03–.07). Also, symptoms and cognitive distortions tended to fluctuate together (bs=.08–.41). The current findings suggest that cognitive distortions do not precede PTSD symptoms as many theories have assumed, but may instead be a consequence or correlate of them. If replicated, our results could help refine cognitive and related models of PTSD.
- Open Access
- 5 June 2026
- Published by Center for Open Science in PsyArXiv
Using LLMs to Evaluate the Interplay Between PTSD Symptom Severity and Intolerance to Uncertainty: A Large-scale EMA Study
Intolerance of uncertainty (IU) is central to cognitive models of posttraumatic stress disorder (PTSD), reflecting a maladaptive process that maintains hypervigilant threat appraisals and symptom persistence. Traditional IU self-report assessments, however, are constrained to explicit, trait-like evaluations and may not capture IU’s implicit and context-dependent nature. Ecological momentary assessment (EMA) combined with large language models (LLMs) offers a scalable approach for detecting subtle linguistic markers of IU in daily life. Here, we introduce a novel language-based method for monitoring IU through naturalistic daily diaries and examine its role in perpetuating PTSD symptoms. In a 90-day EMA study of 432 trauma-exposed individuals with elevated PTSD symptoms (29,623 diary entries with 32,941 daily PTSD symptom ratings), we developed an LLM-based IU assessment, yielding the LLM-based similarity between daily transcribed speech and established IU scale language (r = .60 between LLM measure and human ratings). Using dynamic structural equation modeling (DSEM), we found that higher daily IU predicted greater next-day PTSD symptom severity (β = 0.041, 95% CI [0.024, 0.056]), even after controlling for prior-day symptoms. Conversely, higher daily PTSD symptom severity predicted greater next-day IU (β = 0.035, 95% CI [0.018, 0.054]). The strength of these bidirectional effects varied across individuals (IU→PTSD variance = 0.014, 95% CI [0.007, 0.022]; PTSD→IU variance = 0.037, 95% CI [0.030, 0.047]), reflecting substantial heterogeneity in daily IU–PTSD dynamics. Finally, individuals with stronger IU persistence (IUₜ₋₁ → IUₜ) tended to exhibit enhanced same-day IU–PTSD associations (IUₜ ↔ PCLₜ; r = 0.23, 95% CI = [0.016, 0.410]). In contrast, greater PTSD persistence (PCLₜ₋₁ → PCLₜ) covaried negatively with both lagged (IUₜ₋₁ → PCLₜ; r = -0.38, 95% CI = [-0.505, -0.246]) and contemporaneous (IUₜ ↔ PCLₜ; r = -0.18, 95% CI = [-0.327, -0.024]) IU–PTSD effects, suggesting divergent dynamic patterns relative to IU. Complementary multilevel analyses demonstrated that language-derived IU provides incremental validity, explaining variance in PTSD symptom severity beyond sentiment (NRCSent) and positive and negative affect (LIWC). These findings extend cognitive models of PTSD with IU as a dynamic mechanism influencing symptom trajectories in daily life and demonstrate an ecologically valid framework for tracking implicit cognitive processes via natural language.
- 5 June 2026
- Published by Center for Open Science in PsyArXiv
Open AccessUsing LLMs to Evaluate the Interplay Between PTSD Symptom Severity and Intolerance to Uncertainty: A Large-scale EMA Study
Intolerance of uncertainty (IU) is central to cognitive models of posttraumatic stress disorder (PTSD), reflecting a maladaptive process that maintains hypervigilant threat appraisals and symptom persistence. Traditional IU self-report assessments, however, are constrained to explicit, trait-like evaluations and may not capture IU’s implicit and context-dependent nature. Ecological momentary assessment (EMA) combined with large language models (LLMs) offers a scalable approach for detecting subtle linguistic markers of IU in daily life. Here, we introduce a novel language-based method for monitoring IU through naturalistic daily diaries and examine its role in perpetuating PTSD symptoms. In a 90-day EMA study of 432 trauma-exposed individuals with elevated PTSD symptoms (29,623 diary entries with 32,941 daily PTSD symptom ratings), we developed an LLM-based IU assessment, yielding the LLM-based similarity between daily transcribed speech and established IU scale language (r = .60 between LLM measure and human ratings). Using dynamic structural equation modeling (DSEM), we found that higher daily IU predicted greater next-day PTSD symptom severity (β = 0.041, 95% CI [0.024, 0.056]), even after controlling for prior-day symptoms. Conversely, higher daily PTSD symptom severity predicted greater next-day IU (β = 0.035, 95% CI [0.018, 0.054]). The strength of these bidirectional effects varied across individuals (IU→PTSD variance = 0.014, 95% CI [0.007, 0.022]; PTSD→IU variance = 0.037, 95% CI [0.030, 0.047]), reflecting substantial heterogeneity in daily IU–PTSD dynamics. Finally, individuals with stronger IU persistence (IUₜ₋₁ → IUₜ) tended to exhibit enhanced same-day IU–PTSD associations (IUₜ ↔ PCLₜ; r = 0.23, 95% CI = [0.016, 0.410]). In contrast, greater PTSD persistence (PCLₜ₋₁ → PCLₜ) covaried negatively with both lagged (IUₜ₋₁ → PCLₜ; r = -0.38, 95% CI = [-0.505, -0.246]) and contemporaneous (IUₜ ↔ PCLₜ; r = -0.18, 95% CI = [-0.327, -0.024]) IU–PTSD effects, suggesting divergent dynamic patterns relative to IU. Complementary multilevel analyses demonstrated that language-derived IU provides incremental validity, explaining variance in PTSD symptom severity beyond sentiment (NRCSent) and positive and negative affect (LIWC). These findings extend cognitive models of PTSD with IU as a dynamic mechanism influencing symptom trajectories in daily life and demonstrate an ecologically valid framework for tracking implicit cognitive processes via natural language.
- Open Access
- 1 June 2026
- Published by Center for Open Science in PsyArXiv
Autism, ADHD, and a Neurodevelopmental Spectrum in Hierarchical Taxonomy of Psychopathology (HiTOP)
Autism frequently co-occurs with other mental health conditions, including attention-deficit/hyperactivity disorder. The Hierarchical Taxonomy of Psychopathology (HiTOP) accounts for co-occurrence among mental health conditions by characterizing them using higher-order dimensions, instead of discrete categories. Yet, autism and related neurodevelopmental features are largely missing from HiTOP. We investigated factor analytic models of autism alongside broad psychopathology using Simons Foundation Powering Autism Research for Knowledge, which included autistic youths with sufficient data in preschool (ages 2-5; n=613), childhood (ages 6-11; n=2,703), and adolescence (ages 12-17; n=1,424). A neurodevelopmental spectrum distinct from other HiTOP spectra was identified, with narrower factors representing behavioral rigidity, executive dysregulation, motor, and social communication. The factor structure was consistent across groups, except executive dysregulation was not observed in preschool. Notably, ADHD items did not join the neurodevelopmental spectrum in preschool, loading on externalizing. Findings support inclusion of a neurodevelopmental spectrum in HiTOP. Scientific and clinical implications are discussed.
- 1 June 2026
- Published by Center for Open Science in PsyArXiv
Open AccessAutism, ADHD, and a Neurodevelopmental Spectrum in Hierarchical Taxonomy of Psychopathology (HiTOP)
Autism frequently co-occurs with other mental health conditions, including attention-deficit/hyperactivity disorder. The Hierarchical Taxonomy of Psychopathology (HiTOP) accounts for co-occurrence among mental health conditions by characterizing them using higher-order dimensions, instead of discrete categories. Yet, autism and related neurodevelopmental features are largely missing from HiTOP. We investigated factor analytic models of autism alongside broad psychopathology using Simons Foundation Powering Autism Research for Knowledge, which included autistic youths with sufficient data in preschool (ages 2-5; n=613), childhood (ages 6-11; n=2,703), and adolescence (ages 12-17; n=1,424). A neurodevelopmental spectrum distinct from other HiTOP spectra was identified, with narrower factors representing behavioral rigidity, executive dysregulation, motor, and social communication. The factor structure was consistent across groups, except executive dysregulation was not observed in preschool. Notably, ADHD items did not join the neurodevelopmental spectrum in preschool, loading on externalizing. Findings support inclusion of a neurodevelopmental spectrum in HiTOP. Scientific and clinical implications are discussed.
- Open Access
- 31 May 2026
- Published by Center for Open Science in PsyArXiv
Discriminant Validity: Disentangling Health and Emotional Constructs from Language-Based Assessments
Language-based assessments of health have demonstrated validity in terms of convergence with self-report questionnaire scores, but recent work has shown they lack an important psychometric property: discriminant validity, the ability of the measure to distinguish the target construct from a related one. For example, a language-based general mental health assessment may accurately assess the degree of mental health, but it may also produce scores that correlate with general physical health well beyond theoretical expectations. Here, we propose and evaluate two methods for altering the standard loss function to directly penalize off-target correlations while retaining convergent validity. Evaluated across two clinical language datasets spanning physical and mental health, the augmentations substantially reduced off-target correlations with minimal loss in convergent validity (on-target correlations). Limiting loss in convergent validity to 0.005 Pearson correlation points, the loss function using Squared Cosine Similarity Discrimination significantly improved discriminant validity of language-based mental and physical health assessments (r = 0.454 → r = 0.322; p < 0.05); and fundamental psychopathology dimensions (r = 0.442 → 0.430; p < 0.05). These findings demonstrate that enforcing discriminant validity as a training objective is effective, moving language-based assessments closer to the specificity required for differential clinical use.
- 31 May 2026
- Published by Center for Open Science in PsyArXiv
Open AccessDiscriminant Validity: Disentangling Health and Emotional Constructs from Language-Based Assessments
Language-based assessments of health have demonstrated validity in terms of convergence with self-report questionnaire scores, but recent work has shown they lack an important psychometric property: discriminant validity, the ability of the measure to distinguish the target construct from a related one. For example, a language-based general mental health assessment may accurately assess the degree of mental health, but it may also produce scores that correlate with general physical health well beyond theoretical expectations. Here, we propose and evaluate two methods for altering the standard loss function to directly penalize off-target correlations while retaining convergent validity. Evaluated across two clinical language datasets spanning physical and mental health, the augmentations substantially reduced off-target correlations with minimal loss in convergent validity (on-target correlations). Limiting loss in convergent validity to 0.005 Pearson correlation points, the loss function using Squared Cosine Similarity Discrimination significantly improved discriminant validity of language-based mental and physical health assessments (r = 0.454 → r = 0.322; p < 0.05); and fundamental psychopathology dimensions (r = 0.442 → 0.430; p < 0.05). These findings demonstrate that enforcing discriminant validity as a training objective is effective, moving language-based assessments closer to the specificity required for differential clinical use.
- Open Access
- 18 March 2026
- Published by Center for Open Science in PsyArXiv
Distinguishing the Language of Mental and Physical Health: A Sequential Evaluation with Model Preregistration of Automated Clinical Visit Interviews
Background. Mental and physical health are correlated, yet are described differently in natural language. Standard symptom checklists may limit individuals from fully describing how they experience health problems. This study examines the distinction between how individuals linguistically describe mental and physical health.Method. We used the Sequential Evaluation with Model Pre-registration (SEMP) design to analyze language from automated, open-ended clinical interviews with 9/11 first responders. We developed language-based models assessing Mental and Physical Component Summary (M/PCS) scores from automated interview transcripts (N=1290). We then evaluated these models on held-out prospective data (N = 310) against M/PCS, medical record diagnoses, and healthcare expenditures.Results. Language-based assessment correlated significantly with M/PCS scores (r = .36 and r = .37, respectively, all p<.001), exceeding SEMP pre-registered thresholds (r>.315, and r>.348, respectively). Analyses revealed both shared and distinct linguistic markers for mental and physical health, with language reflecting emotional, somatic, and functional themes. The language-based assessments demonstrated external validity through significant convergence with clinical diagnoses (AUCs=.74 and .65 for mental and physical health, respectively). Furthermore, language-based assessments exhibited incremental validity; when added to models containing traditional self-reports, they explained significant unique variance in both mental (ΔR²=.063, p<.001) and physical (ΔR²=.032, p=.010) healthcare costs.Conclusions. Natural language from automated interviews captures general health through emotional, somatic, and functional themes. Validated against clinical diagnoses and healthcare costs, these models demonstrate that open-ended language provides interpretable insights beyond traditional rating scales, potentially augmenting clinical evaluations and understanding.
- 18 March 2026
- Published by Center for Open Science in PsyArXiv
Open AccessDistinguishing the Language of Mental and Physical Health: A Sequential Evaluation with Model Preregistration of Automated Clinical Visit Interviews
Background. Mental and physical health are correlated, yet are described differently in natural language. Standard symptom checklists may limit individuals from fully describing how they experience health problems. This study examines the distinction between how individuals linguistically describe mental and physical health.Method. We used the Sequential Evaluation with Model Pre-registration (SEMP) design to analyze language from automated, open-ended clinical interviews with 9/11 first responders. We developed language-based models assessing Mental and Physical Component Summary (M/PCS) scores from automated interview transcripts (N=1290). We then evaluated these models on held-out prospective data (N = 310) against M/PCS, medical record diagnoses, and healthcare expenditures.Results. Language-based assessment correlated significantly with M/PCS scores (r = .36 and r = .37, respectively, all p<.001), exceeding SEMP pre-registered thresholds (r>.315, and r>.348, respectively). Analyses revealed both shared and distinct linguistic markers for mental and physical health, with language reflecting emotional, somatic, and functional themes. The language-based assessments demonstrated external validity through significant convergence with clinical diagnoses (AUCs=.74 and .65 for mental and physical health, respectively). Furthermore, language-based assessments exhibited incremental validity; when added to models containing traditional self-reports, they explained significant unique variance in both mental (ΔR²=.063, p<.001) and physical (ΔR²=.032, p=.010) healthcare costs.Conclusions. Natural language from automated interviews captures general health through emotional, somatic, and functional themes. Validated against clinical diagnoses and healthcare costs, these models demonstrate that open-ended language provides interpretable insights beyond traditional rating scales, potentially augmenting clinical evaluations and understanding.
- Open Access
- 26 February 2025
- Published by Center for Open Science in PsyArXiv
Day-to-day dynamics of facial emotion expressions in posttraumatic stress disorder
Facial expressions are an essential component of emotions that may reveal mechanisms maintaining posttraumatic stress disorder (PTSD). However, most research on emotions in PTSD has relied on self-reports, which only capture subjective affect. The few studies on outward emotion expressions have been hampered by methodological limitations, including low ecological validity and failure to capture the dynamic nature of emotions and symptoms. Our study addresses these limitations with an approach that has not been applied to psychopathology: person-specific models of day-to-day facial emotion expression and PTSD symptom dynamics. We studied a sample of World Trade Center responders (N=112) with elevated PTSD pathology who recorded a daily video diary and self-reported symptoms for 90 days (8,953 videos altogether). Facial expressions were detected from video recordings with a facial emotion recognition model. In data-driven, idiographic network models, most participants (80%) had at least one, reliable expression-symptom link. Six expression-symptom dynamics were significant for >10% of the sample. Each of these dynamics had statistically meaningful heterogeneity, with some people’s symptoms related to over-expressivity and others to under-expressivity. Our results provide the foundation for a more complete understanding of emotions in PTSD that not only includes subjective feelings but also outward emotion expressions.
- 26 February 2025
- Published by Center for Open Science in PsyArXiv
Open AccessDay-to-day dynamics of facial emotion expressions in posttraumatic stress disorder
Facial expressions are an essential component of emotions that may reveal mechanisms maintaining posttraumatic stress disorder (PTSD). However, most research on emotions in PTSD has relied on self-reports, which only capture subjective affect. The few studies on outward emotion expressions have been hampered by methodological limitations, including low ecological validity and failure to capture the dynamic nature of emotions and symptoms. Our study addresses these limitations with an approach that has not been applied to psychopathology: person-specific models of day-to-day facial emotion expression and PTSD symptom dynamics. We studied a sample of World Trade Center responders (N=112) with elevated PTSD pathology who recorded a daily video diary and self-reported symptoms for 90 days (8,953 videos altogether). Facial expressions were detected from video recordings with a facial emotion recognition model. In data-driven, idiographic network models, most participants (80%) had at least one, reliable expression-symptom link. Six expression-symptom dynamics were significant for >10% of the sample. Each of these dynamics had statistically meaningful heterogeneity, with some people’s symptoms related to over-expressivity and others to under-expressivity. Our results provide the foundation for a more complete understanding of emotions in PTSD that not only includes subjective feelings but also outward emotion expressions.
- Open Access
- 4 October 2024
- Published by Center for Open Science in PsyArXiv
Replicability and Validity of a New AI Assessment of PTSD from Patient Language: A Sequential Evaluation with Model Pre-registration
AI shows promise in identifying psychopathology through language, but replicability in AI-models remains challenging. We develop a AI-based language assessment of PTSD severity and introduce the Sequential Evaluation with Model Pre-registration to rigorously evaluate its validity and replicability. This design includes two phases: Development with pre-registration and Evaluation. Data included development (N=1437) and prospective (N=346) samples, where participants described their lives during automated interviews. In the prospective sample, pre-registered models correlated with PTSD CheckList scores (r=.38, p<.001); converged with PTSD diagnosis (AUC=.76; outperforming demographics and trauma exposures: AUC=.61, p<.01). We found that for each standard deviation increase, mental healthcare expenditure rose by $696.5 (p<.001). Our pre-registered PTSD model assessments are replicable in prospectively collected clinical data and showed external validity against expense criteria. With further development, such models can be used to screen for PTSD or monitor treatment response, especially in telehealth or automated interviews, where deployment can be seamless.
- 4 October 2024
- Published by Center for Open Science in PsyArXiv
Open AccessReplicability and Validity of a New AI Assessment of PTSD from Patient Language: A Sequential Evaluation with Model Pre-registration
AI shows promise in identifying psychopathology through language, but replicability in AI-models remains challenging. We develop a AI-based language assessment of PTSD severity and introduce the Sequential Evaluation with Model Pre-registration to rigorously evaluate its validity and replicability. This design includes two phases: Development with pre-registration and Evaluation. Data included development (N=1437) and prospective (N=346) samples, where participants described their lives during automated interviews. In the prospective sample, pre-registered models correlated with PTSD CheckList scores (r=.38, p<.001); converged with PTSD diagnosis (AUC=.76; outperforming demographics and trauma exposures: AUC=.61, p<.01). We found that for each standard deviation increase, mental healthcare expenditure rose by $696.5 (p<.001). Our pre-registered PTSD model assessments are replicable in prospectively collected clinical data and showed external validity against expense criteria. With further development, such models can be used to screen for PTSD or monitor treatment response, especially in telehealth or automated interviews, where deployment can be seamless.